A current use of AI is where the system was developed on time, on budget, and met all the technology acceptance criteria, but after implementation, there are new concerns raised that had never been considered in the initial plan, such as:
Requirements regarding when and how users should use the technology.
--Who owns the AI decisions? Who controls the updates to the models? What if the outputs can’t be explained?
These concerns are no longer in the realm of theory. As it stands, AI is now a part of the delivery cycle and can't be relegated to the other side of the project life cycle, after the fact.
From the project management point of view, the following challenges arise:
--Where do AI Ethics & Governance Checkpoints relate to traditional Stage Gates?
--How do we deal with risks like bias, regulation, and accountabilities that emerge as a consequence of changing model requirements?
--What does “done” mean for an AI-infused deliverable, anyway?
It seems that a new role for project managers is emerging as a conduit between data teams, business, the law department, and risk organization. This shifting role that I watch happening in real-time would be interesting to hear about in terms of how you're actually incorporating AI governance and ethics into day-to-day project delivery.
You raise a significant question about the current gaps between traditional IT, data governance and ethics practices for AI projects.
I have been studying AI for 3+ years with an increasing focus on Responsible AI and Explainable AI and believe that AI technology is moving so quickly that AI ethics and governance are struggling to keep up, which presents serious business risks.
AI ethics and governance are key foundational components that need to be part of Gen AI projects from the very beginning to ensure training/RAG and proprietary data are ethically sourced, adequately protected, and managed responsibly. An emphasis must be placed on protecting PII, PHI, IP and other secret data at all on-prem and/or cloud locations and stages of the project to prevent unintended leakage and protect it from hackers as malicious attacks are growing and increasing in complexity.
AI ethics and governance are needed at every stage of the project, which means data scientists, data engineers, ML engineers, 3rd party vendors, domain experts, end users, etc. must be trained effectively and aligned to the policies, procedures, incident reporting, and remediation.
AI ethics and governance teams must be sufficiently knowledgeable about data and AI to properly identify the unique risks associated with AI and work in close partnership with business leaders and technical teams. Likewise, project managers should educate themselves to better support their AI projects.
I see this as a space that will continue to grow and expand as AI moves deeper into business systems and our lives in general. Saving Changes...
Program Manager| HARPER SRLSanto Domingo / Distrito Nacional, Dominican Republic
AI can meet delivery criteria and still surface governance, ethical, and accountability issues only after deployment. That forces project managers to extend their role beyond delivery into coordination across legal, risk, data, and business teams. Treating ethics and governance as ongoing delivery concerns, not end-stage reviews, is becoming essential. “Done” for AI now includes explainability, ownership clarity, and controls that continue after go-live. Saving Changes...
With AI-enabled projects, meeting scope, schedule, and technical acceptance is no longer enough. Governance, accountability, and usage risks often emerge after go-live, which means ethics and controls need to be embedded into the delivery lifecycle, not added later.
What has worked for me is treating AI governance as a continuous check aligned to stage gates, and redefining “done” to include ownership, guardrails, and monitoring - not just deployment.
The PM role is clearly shifting into a connector between delivery teams, risk, legal, and the business. Saving Changes...
Luis BrancoCEO| Business Insight, Consultores de Gestão, LdªCarcavelos, Lisboa, Portugal
I agree with the core diagnosis. The issue is no longer AI arriving late in the project lifecycle, it is governance arriving too late.
In AI-enabled projects, “done” cannot mean technical delivery alone. A model that works but has unclear rules for use, explainability, updates, and accountability is not finished, it is merely live. Acceptance becomes a matter of governance and ethics, not just functionality.
Ethics and governance checkpoints also do not fit neatly into traditional stage gates, because they are not one-off events. Bias, model drift, regulatory change, and new usage patterns often emerge after go-live. That requires continuous governance, not retrospective control.
The project manager’s role is clearly shifting. Less a manager of artifacts, more an integrator of responsibility across business, data, risk, legal, and human impact. If governance does not move at the same pace, AI will keep delivering faster than organizations can decide with awareness. That is a management risk, not a technology one. Saving Changes...
Sergio Luis ConteHelping to create solutions for everyone| Worldwide based OrganizationsBuenos Aires, Argentina
This is a big fallacy in the statement itself: AI Delivers Faster Saving Changes...
Tim McClungProject Coordinator| New Hanover County ITWilmington, Nc, United States
I see this gap clearly. AI can be delivered on time and on budget, yet still raise governance and accountability issues after go-live.
I treat AI ethics and governance as continuous checkpoints, not one-time stage gates. For me, “done” means the solution is usable, explainable, and has clear ownership.
The PM role is evolving into a connector between data, business, legal, and risk, early and often. Saving Changes...
Community Champion
